Sign language

How can a full language exist with no sound in it at all?

Ask most hearing people what a sign language is and you will get some version of: gestures standing in for the words of the local spoken language. That guess makes a testable prediction — that Britain and the United States, sharing a spoken language, should share a sign language. They do not. British Sign Language and American Sign Language are mutually unintelligible; ASL is historically related to French Sign Language, because a French teacher helped found the first American deaf school. The guess fails at the first check. What, then, are these languages made of?

Start with the piece that seems impossible. Spoken languages are built from a small inventory of meaningless units — consonants and vowels — combined into meaningful words, and that double layering is what lets a few dozen sounds yield tens of thousands of words. If a language has no sound, does it lose the lower layer? Are signs simply whole gestures, each an unanalysable picture?

That was the assumption until William Stokoe, in 1960, tried decomposing ASL signs anyway. He found a sign to be not one indivisible act but a simultaneous bundle of choices: what shape the hand takes, where it is placed, how it moves — with orientation and facial markers added by later work. Are those choices really contrastive, the way /p/ and /b/ are in English? Test it as a linguist would, by hunting minimal pairs. In ASL, MOTHER and FATHER share a handshape and a movement and differ only in location, at the chin or the forehead. A commonly cited trio of otherwise identical signs at forehead, nose and chin makes the point three ways. Change one parameter and you have a different word, not a variant pronunciation.

There is behavioural evidence too. Fluent signers make slips of the hand, substituting one sign's handshape into another exactly as speakers swap consonants between words — and you cannot slip on a unit that does not exist. So the lower, meaningless layer is there; it is simply not made of sound. Linguists still call this level phonology, an odd name kept for a real reason: the level does the same structural work regardless of channel.

Now the grammar. If signing were pantomime, the space in front of the signer would be scenery. Instead it is machinery: a signer establishes a referent at a point in space and then moves a verb from one point to another to mark who did what to whom — agreement done geometrically. Facial behaviour is not emotional colouring but grammar; in ASL a raised brow marks a yes/no question, a different brow configuration a wh-question, a head shake negation, and these run simultaneously with the hands, a genuine design difference from speech rather than a version of it.

Is this really the same faculty as spoken language, or something spatial pretending to be one? The neurological evidence is unusually clean. Damage to the classic left-hemisphere language areas produces aphasia in signers — disrupted signing with intact non-linguistic spatial ability — while right-hemisphere damage can wreck spatial cognition and leave the grammar working. The brain files signing under language, not gesture.

That leaves the deepest question: where does the grammar come from, if nobody teaches it? Nicaragua produced something close to an experiment. Deaf Nicaraguans were largely isolated from one another until schools opened in Managua in 1977 and 1980. Instruction was in spoken Spanish and lipreading, and it largely failed — but the children, together in numbers for the first time, pooled their home gesture systems in the playground and the buses. The first cohort produced a shared but variable system. Then younger children arrived, took that system as their input, and produced something more regular than what they had received, using devices their models did not consistently use — separating the manner and the path of a motion into distinct signs, for one. The language got more systematic as it passed to the children learning it, not less.

Three refinements are worth holding onto. Iconicity is real but not disqualifying: many signs do resemble what they mean, the resemblance conventionalises over time, and the grammar around them stays arbitrary — spoken languages have their own iconic pockets and nobody calls them pantomime.

A sign language is also not a manual encoding of a spoken one. Systems that do encode a spoken language onto the hands exist — fingerspelling, and signed versions of English devised for classrooms — and they are deliberately unlike natural sign languages, which have their own word order, morphology and idioms.

And the Nicaraguan case is evidence about acquisition, not proof of an innate grammar module. What it shows firmly is that children given a partial, irregular system regularise it: the structure appeared in the learners rather than the input. What that implies about the mind is still argued.

• Homesign: the structured systems deaf children invent alone, before any community exists.
• How simultaneity changes what a grammar can do — what sign languages express that speech cannot easily.
• Why iconicity erodes as signs conventionalise, and what that predicts about young sign languages.

Every term the collection defines is gathered in the glossary.